Recent Research Projects

Michael Lemmon, University of Notre Dame

• Using Data Science to Protect Tap Water Quality
  Lucy Family Institute for Data and Society, March 30 2022-2023
  
  OVERVIEW: The proposed research uses data science to identify homes at risk for unhealthy tap water and develop data science-based strategies to mitigate these risks. While tap water is safe in the majority of homes in the US, there are notable exceptions, as seen recently in Flint, MI and Benton Harbor, MI. Unfortunately, water quality problems often impact low-income families, which are the least prepared to manage them. In most cases, water supplied to homes from the public distribution networks meets stringent EPA drinking water standards. The problems occur within homes. Pipes may leach toxic metals, such as lead and copper, or promote growth of pathogenic bacteria, such as Legionella pneumophila. In both cases, the decay in water quality correlates to the “water age” or stagnation within the plumbing system. Longer water ages provide more time for metals to leach and for bacteria to grow. Given the complex nature of plumbing systems that 1) may have been built over 100 years ago and lack documentation, 2) may have been modified over the years without documentation, 3) have pipes that are often hidden beneath the ground or behind walls, and 4) have highly variable water demands whose characteristics change depending on the type of fixtures and appliances, and the number and type of occupants, home plumbing systems often behave as black boxes. Data science can be used to identify which homes are at risk, and to develop strategies to mitigate the risks. For example, actively controlling water age in premise plumbing systems could help improve water quality and reduce health risks. The specific objectives of this project are to (1) develop a GIS-based model to predict homes with high water ages, and (2) develop machine-learning-based strategies to reduce water age. This project studies a data-based approach to identify homes with potential water quality concerns:, to identify homes likely to have high water ages.

---

• CPS: SMALL: Learning How to Control - A Meta-Learning Approach for the Adaptive Control of Cyber-Physical Systems
  National Science Foundation, June 15, 2023 - June 14, 2026
  
  OVERVIEW: Internet-of-Things (IoT) enabled manufacturing systems (a.k.a. Manufacturing 4.0) consist of two networks. There is a physical network formed from the machines carrying and processing materials on the factory floor. There is also a cyber network formed from the wired and wireless communication networks over which factory machines exchange information. Unexpected disturbances on the factory floor interrupt the flows in both networks and this congestion introduces a degree of uncertainty that has been an obstacle to the adoption of IoT-enabled technologies by U.S. manufacturers. This project removes that obstacle through machine learning algorithms that learn how to control the physical and cyber networks of an IoT-enabled manufacturing system. ``Learning how to Control'' is a meta-learning approach to control that “learns” how to configure the control problem so one can synthesize a base control policy whose performance is robust to unexpected changes in the system's task environment. This approach to adaptive control will be evaluated on a multi-robotic testbed mimicking the use of WIFI connected robots moving materials across a factory floor. This project will take the models and policies learned on the testbed and investigate methods that transfer them to IoT-enabled factories found in local manufacturing facilities.